quality of life and the migration of the college-educated ...naumannj/professional geography...

Quality of Life and the Migration of theCollege-Educated: A Life-Course Approach

RONALD L. WHISLER, BRIGITTE S. WALDORF,GORDON F. MULLIGAN, AND DAVID A. PLANE

ABSTRACT This paper examines how the college-educated population—segmented into selec-

tive demographic groups, from young adults to the elderly—differentially values quality-of-life

(QOL) indicators of metropolitan areas in the United States. Using data from the 2000 Census and

the 1997 Places Rated Almanac, out-migration patterns are shown to depend jointly upon stage in

the life course, the spatial-demographic setting, and QOL characteristics. An abundance of cultural

and recreational amenities lowers out-migration rates of young college-educated. For the older

college-educated population, the revealed preferences shift toward concerns for safety and a strong

preference for milder climates. The study also finds significantly lower out-migration rates for

metropolitan areas with growing human capital. In light of shifting age distributions and rising

educational attainment levels, the results have important implications for the emergence of new

migration patterns and the concentration of human capital.

Introduction

O ne of the most popular rationales for migration is Tiebout’s observation that people“vote with their feet.” It is now widely recognized that this “migration vote” is

contingent on personal characteristics including life-course attributes and on location-specific characteristics including natural and man-made amenities. Amenities are notmeasures of overall residential desirability of places, but rather place-specific attributes thatpeople differentially value at different stages of their life. It is also critical to recognize thatplaces offer bundles of amenities: consequently migrants must sometimes take the badwith the good in choosing the place where its set of attributes best satisfies their tastesand preferences.

Ronald L. Whisler is a statistical analyst at The Modellers, LLC, in Salt Lake City, UT. His email

address is [email protected]. Brigitte S. Waldorf is a professor in the Department of Agricultural

Economics at Purdue University in West Lafayette, Indiana. Gordon F. Mulligan is a professor

emeritus and David A. Plane is a professor in the Department of Geography and Regional Develop-

ment at the University of Arizona in Tucson, Arizona. The authors are grateful to Carmen Carrión-

Flores, Jacques Poot, and Todd Sorensen for their helpful comments, to Henry W. Herzog, Jr. for his

advice on the research design, and to L. Benjamin Luzynski for his assistance with GIS. The authors

thank the editor and three anonymous reviewers for their excellent comments and suggestions.

Growth and ChangeVol. 39 No. 1 (March 2008), pp. 58–94

Submitted August 2006; revised June 2007, September 2007; accepted October2007.© 2008 Blackwell Publishing, 350 Main Street, Malden MA 02148 US and 9600Garsington Road, Oxford OX4, 2DQ, UK.

mailto:[email protected]

Understanding the interplay of personal characteristics and amenities in migrationdecisions is pivotal not only because valuations of amenities may differ across life-coursegroups, but also because of anticipated changes in the composition of the U.S. population.Trends such as the retirement of the baby boomers and the emergence of new demographicgroups like the “power couples” (Costa and Kahn 2000) will lead to shifts in the relativeimportance of different demographic groups. Thus, demographic changes in combinationwith differential valuations of quality-of-life (QOL) characteristics may trigger substantialshifts in migration patterns and, ultimately, in the nation’s population distribution.

To address the joint impact of personal characteristics and amenities on the “migrationvote,” this paper revisits the seminal U.S. migration study by Herzog and Schlottmann(1986). Their study was particularly important because out-migration patterns from U.S.Metropolitan Statistical Areas (MSAs) were clearly shown to depend both on the charac-teristics of households and on how those households assessed inter-city differences in QOLindicators. Our research takes the Herzog and Schlottmann study a step further by showinghow the valuation of those amenities differs across life-course groups. Toward this end,we use a demographic segmentation approach that allows stratification of our householdsample and—as suggested by Cushing (1993)—enables us to estimate separate out-migration models for several life-course groups. To the extent that amenities are differen-tially valued across life-course groups, migration models estimated without demographicdisaggregation are misspecified in positing general effects of various place-specificamenities.

In total, we distinguish between six life-course groups that reflect very different stagesin the life course. Age, marital status, and presence of children distinguish these separatedemographic groups. Moreover, for each life-course group, we concentrate solely on themigration behavior of the college-educated population. Three reasons justify this selection.First, the college-educated population in the United States has very high migrationpropensities (Kodrzycki 2001) and is thus one of the most influential driving forces forinternal population redistributions. Second, in today’s knowledge-based economy, thecollege-educated constitute the most sought-after segment of the population. A well-educated population positively influences an area’s competitiveness and losing the college-educated eventually takes a toll on local economies. Thus, understanding the role ofamenities—frequently postulated to be the sine qua non for a strong presence of whatFlorida (2002b) calls the “creative class”—in enhancing a place’s retention power is offoremost importance. Finally, from a methodological perspective, a focus on the college-educated ensures a certain comparability by income or, at least, by income potential.

The empirical analysis draws on the site-specific amenity scores given in the 1997Places Rated Almanac and other data on metropolitan attributes and personal characteris-tics made available from the 2000 U.S. Census. For each life-course group, out-migrationpropensities are specified as a function of amenity scores and a series of control variables.A logit framework is used for estimation.

The paper is organized as follows. In the next section, we provide a brief overview of theliterature connecting amenities, life cycle, and migration. The third section revisits and

QUALITY OF LIFE AND MIGRATION OF COLLEGE-EDUCATED 59

updates the important Herzog and Schlottmann (1986) study. The fourth section describesthe research design, paying special attention to the QOL variables, the stratification of thesix life-course groups, and the model specification. The empirical results are presented inthe fifth section. The paper concludes with a summary of findings and future researchdirections.

Previous ResearchThis study was informed by several overlapping research streams. One of these was the

literature devoted to the postindustrial landscape, where analysts have increasingly recog-nized the role played by amenities or QOL factors in the location decisions of householdsand firms (Dissart and Deller 2000; Mulligan, Carruthers, and Cahill 2004). Other ideaswere drawn from the very wide literature on household migration, where some analystsprefer to assess aggregate movements of people while others identify those diverse factorsaccounting for the choices made by individual movers (Greenwood 1975; Greenwood,Mueser, et al. 1991). A third research stream focuses more on the decision-making of thehousehold and attempts to uncover those internal attributes and external forces that dictatebehavioral change at certain key points in the life course (Golledge and Stimson 1997).Finally, the study is embedded in the quickly growing literature on locational decisions andspatial concentrations of the highly educated population (Adamson, Clark, and Partridge2004; Gottlieb and Joseph 2006).

Amenities. Amenities are site- or region-specific goods and services that make locali-ties more or less attractive to agents. They enter into the consumption decisions of house-holds and the production decisions of firms, and therefore, play an important role in thelocation choices of these agents (Diamond and Tolley 1982). Amenities first became ofwide interest with the emergence of postindustrialism (Bell 1973), particularly in theUnited States (Berry and Horton 1970). They were originally considered to be largelynatural but now man-made amenities and disamenities are of greater interest for reasons ofpublic policy (Carruthers and Mundy 2006). Today, there is widespread recognition thatman-made amenities affect both the level and the distribution of society’s scarce resourcesthrough spatial externalities (Smith 1977), a geographic fact that is now seen most clearlyin the world’s very largest cities (Scott 2001).

Much research is driven by the notion of compensating differentials, which followsfrom Rosen (1974) and Roback (1982). Here location-specific amenities work through bothlabor and land markets, so representative households can be indifferent between livingin attractive (low wages, high rents) versus unattractive places (high wages, low rents).Among large metropolitan areas, this compensation process appears to work at both inter-and intra-regional levels, as evidenced by Bloomquist, Berger, and Hoehn (1988). Researchhas extended compensating differentials to local fiscal conditions as well and well-managedlocal governments affect the behavior of firms and households much like valued environ-mental amenities (Gyourko and Tracy 1991). In more recent times, people like Glaeser(1999), Costa and Kahn (2000), and Florida (2002a) have argued that young, well-educatedhouseholds have strong preferences for certain kinds of urban milieus in order to satisfy

60 GROWTH AND CHANGE, MARCH 2008

their lifestyle demands and maximize their life chances. Invariably, these milieus include,but are not restricted to, the following list of attributes: wide variety in consumer goods andservices; diverse entertainment and recreational possibilities; dense networks of educa-tional, employment, and social opportunities; and tolerant racial and social attitudes.

Migration. Migration is the location adjustment made by households in response totheir ever-changing needs and preferences. In the United States, some 15 to 20 percent ofthe population change their residence every year, suggesting a nation that is perpetuallyreallocating its human resources in space as some places (e.g., neighborhoods and regions)rise and others decline.

The reasons for these moves are diverse. Relocation occurs in those situations where insitu adjustments fail to bring down high levels of stress to tolerable levels (Brown and Moore1970; Wolpert 1965). Aggregate models of this behavior typically include two types ofattributes: those that either repel households from one place or attract them to anotherand those that affect the ease of movement between those places. From a human capitalperspective (Sjaastad 1962), migration is one of several means by which a household canimprove its life chances. Here the household calculates the present value of place-specificbenefits and the various costs of moving to other places before deciding on staying or moving.Finally, neoclassical theories of regional growth argue that households move in response towage-rate differences, although the interregional gap between nominal and real wages canvary significantly depending upon the geography of amenities. But here it is difficult to gaugethe quality of information that households have about such matters as job openings, housingvacancies, and education opportunities at other locations, all factors that can severely inhibitthe movement of households to those alternative destinations (Isserman 1986).

Neoclassical migration models were devised during the industrial era when workerstended to move from low-wage areas with labor surpluses to high-wage areas with laborshortages (Borts and Stein 1964). This disequilibrium model suggests that regional wagesequilibrate over the long run although some observers pointed out that migration tends tobe both circular and selective (Greenwood 1973; Thirlwall 1966). In any case, (spatial)equilibrium models have been more popular in recent times. As in urban hedonic models,household behavior is seen to be strongly affected by the geography of amenities anddisamenities, so that nominal differences in wages between labor markets, even when theyare fairly substantial, do not necessarily lead to migration.

This approach was pioneered by Graves (1979, 1980, 1983), along with Graves andLinneman (1979), and a flood of later studies repeatedly has confirmed the influence ofamenities on migration (e.g., Clark and Hunter 1992; Clark, Knapp, and White 1996; Clark,Lloyd, and Jain 2002; Porell 1982; Rappaport 2007). The demand for place-specificamenities has price and income effects just like any other good or service, and localizedshifts in demand can be initiated by age, gender, and race changes in the local population.For this paper, a couple of findings are especially worthy of note. First, the approachstrongly endorses the importance of life cycle in hedonic migration research. Clearly, anatural amenity like winter warmth is valued by retirees as a normal good while adisamenity like humidity is valued as an inferior good. Second, significant behavioral


differences exist across subpopulations depending upon the resources they can access.Winter warmth might be valued as a superior good by richer households but only as anormal good by poorer households. Man-made amenities like public education and dis-amenities like crime presumably exhibit the same sorts of price and income effects (Porell1982).

Related work by Clark and Hunter (1992) has indicated that the disequilibriumapproach clearly loses its explanatory ability for older-age cohorts who increasingly valueQOL factors as they approach retirement age. Other research by Greenwood, Hunt, et al.(1991) and Mueser and Graves (1995) has sought to reconcile the disequilibrium andequilibrium approaches over a lengthy time interval, finding that in some periods, firmprofit was dominant in driving job-related migration while in other periods, householdutility was dominant in driving amenity-related migration. In any case, this line of researchindicates that an appreciation of the demographic makeup and socio-economic compositionof each region’s population is not only critical for assessing how jobs and amenities aredifferentially valued in those regions, but is also very important for understanding thenature of household movement that takes place between those regions.

Life course. In demography, sociology, and geography, household life-cycle modelshave been popular for decades in migration studies and in the social area analysis of largemetropolitan areas (Berry and Horton 1970). Today, there is no doubt that stage in thehousehold life cycle is one of the main determinants of residential relocation in theadvanced economies. The postindustrial improvements in health and longevity have otherimportant economic implications worthy of note. First, as Becker (1993) pointed out,people now have a greater incentive to invest in human capital because they can enjoy thebenefits over a longer period of time. This means that individuals have probably becomemore rational about when and where they actually decide to carry out this investment,whether they choose deepening their skills or furthering their education. Second, post-industrial households occupy more life-course stages than their industrial predecessors andtheir longitudinal consumption patterns now change a lot as they pass through life’srecognizable stages. Indeed, Wakabayashi and Hewings (2007) provide recent evidenceof this from Japan, where they suggest that such changes in consumption behavior haveimportant implications for structural change in postindustrial economies. Moreover, theirdata indicate that significant regional variation exists in the budget shares of households, afinding that confirms that entire regions can have very different aggregate consumptionprofiles. An implication of this is that households will relocate—at certain strategictimes—from one region to another in order to take advantage of the place-specific con-sumption opportunities that are more available in that other region.

The sociologist Rossi’s (1955) work was important for connecting the household’slife-cycle stages to its preferences and needs for housing. In subsequent work, the term lifecourse often has been substituted, apparently to avoid age and structural stereotypes duringthis era of postindustrialism (Levinson 1978). Some observers have suggested as manyas ten different life-cycle events that are related to residential adjustment or relocation(Rowland 1982). Completion of tertiary education, marriage, and retirement are widely


acknowledged to be three of the most important events in the life course, although itis acknowledged that the life-cycle factor is a useful predictor of relocation only whenit is combined with other lifestyle and socio-economic information about the household.

There certainly exists a wealth of studies addressing the issue of age-dependency inmigration, a notion that is perhaps most succinctly captured in the concept of migration ageschedules (Pandit 1997; Rogers 1988). The dependency of migration on life-course eventshas been highlighted in a recent study by Plane and Heins (2003), who identified seven agecohorts based on characteristic migration patterns. In other words, knowledge of the originsand destinations of migration streams provides some clues by itself to the life-courseattributes of the households that are actually involved in moving. In addition to age, maritalstatus is an important descriptor of the life course influencing migration behavior. Theseminal work by Mincer (1978), complemented by the studies of Graves and Linneman(1979) and Schaeffer (1987), emphasize the reduced migration propensities of marriedcompared to nonmarried persons. More recent studies (e.g., Jürges 2006; Nivalainen 2004;Swain and Garasky 2007) continue to emphasize the importance of marital status, often incombination with gender roles.

Understanding the joint influence of differential valuation of place attributes across thelife course is even more important when considering the migration patterns of the college-educated. Their high income potential imposes fewer constraints on relocating according totastes and preferences. However, the literature has not yet addressed this issue in adequatedetail. Instead, research on the locational choices of the highly educated focuses eitheron the young as a special group (e.g., Franklin 2003; Gottlieb and Joseph 2006) or on theentire college-educated workforce (e.g., Waldorf 2007).

What we do know with certainty is that the highly educated population is a particularlymobile group. Basker (2002) reports that persons with at least a bachelor’s degree nowaccount for 37 percent of all migrants but they comprise only 27 percent of the population.The highly educated also have very distinct migration patterns. Kodrzycki (2001), forinstance, finds that the college-educated population is disproportionately likely to engagein long-distance moves. Thirty percent of graduates no longer live in the state where theyattended college, and an even greater percentage no longer live in the state where theycompleted high school. These percentages far exceed the figures for those who nevergraduated from college. Recent research by Wilkinson (2007) further shows that thecollege-educated are less likely than those with lower levels of educational attainmentto move downward within the urban hierarchy. There is also evidence that the college-educated population itself becomes an amenity that attracts other well-educated migrants.Waldorf (2007) finds that highly educated migrants are disproportionately attracted by thehigh educational status of current residents, while Gottlieb and Joseph (2006) find thatscience and technology graduates are especially prone to move to those places havingbetter-educated populations. If these selective migration behaviors continue over a pro-longed time period, the spatial distribution of the highly educated population will becomeincreasingly concentrated, possibly creating striking disparities between brain-rich andbrain-poor regions.


Equally important, the literature also contends that human capital propels regionaleconomic growth (Mathur 1999), where many observers argue that amenities play aprominent role in attracting the college-educated workforce (Florida 2000a,b; Glaeser,Kolko, and Saiz 2000). Recently, Gottlieb and Joseph (2006) find that Ph.D. graduates areparticularly responsive to amenities when making their locational choices. Clark, Lloyd,and Jain (2002) even argue that for the college-educated workforce, “the decision aboutwhere to live and enjoy life can play as large or a larger role than the job offer itself in thefinal location decision” (p. 513). While, in general, amenities and their power to attract thehighly educated are increasingly viewed as the new driving force of urban and regionalgrowth, little is known about how amenity valuations of the highly educated may differacross the various stages of the life course.

Updating Herzog and Schlottmann (1986)Herzog and Schlottmann analyzed individual responses to macro site-specific amenity

data, estimating a binary logit model of leaving versus staying in a MSAs. The sample wasrestricted to employed individuals between the ages of twenty-three and sixty, who wereliving in their state of birth in 1975, not attending college, and not serving in the military.

The dependent variable was proxied using the 1980 U.S. Census 1 percent Public-UseMicrodata Sample (PUMS) data on residential location in 1975 and in 1980. Categorizedas movers were individuals who, in 1980, resided anywhere outside the MSAs in whichthey resided in 1975. Personal attributes—age, race, gender, marital status, presence ofchildren, and years of education—were obtained from the 1980 U.S. Census 1 percentPUMS data. Information on place-specific attributes was taken from the 1981 Places RatedAlmanac, where amenity scores were generated to describe metropolitan variations inliving costs, transportation, job opportunities, education, climate, crime, the arts, healthcare, and recreation.

We updated the Herzog and Schlottmann (1986) model using the 1997 Places RatedAlmanac and the 2000 U.S. Census 5 percent PUMS. Table 1 shows the results originallypresented by Herzog and Schlottmann and the updated results. Both studies similarlysuggest that people tend to leave metropolitan areas with high costs of living and high crimerates, whereas rich recreational opportunities dampen out-migration propensities. In thelate 1990s, just like twenty years earlier, we find low propensities to migrate for African-Americans, older persons, poorly educated persons, and households with children. Neithergender nor marital status has a significant influence on migration propensities in either timeperiod.

However, there are also important differences. Unlike in the 1975–1980 time period,out-migration propensities for 1995–2000 are higher for areas with a poor job outlook anda deprived cultural environment, whereas counterintuitive signs are found for the educationand climate variables. The updated model suggests that people are prone to leaving metroareas with a wealth of educational opportunities and an agreeable climate. These counter-intuitive findings may simply reflect high out-migration from fast-growing metro areas


with high in-migration rates. They might also be explainable if QOL amenities are, in fact,valued differently across life-course groups.

Herzog and Schlottmann applied the estimated parameters to U.S. metropolitan areasto rank them by out-migration rates. Table 2 shows the fifty metro areas with the lowest

TABLE 1. UPDATED AND ORIGINAL PARAMETER ESTIMATES OF THE HERZOG AND

SCHLOTTMANN (1986) MODEL.

Updated Original

Constant -0.374594** -0.8338Economic and quality-of-life variables

Housing (cost of living)a -0.002790*** 0.0752***Economics (job outlook)b -0.000851* 0.0442Climate 0.002510*** -0.2503Recreation -0.005760*** -0.2992**Arts -0.003720*** 0.0166*Education 0.002510** -0.2566Health care 0.000543 -0.0769Transportation -0.001320 -0.0200Crimec -0.001880*** 0.2960**

Personal characteristicsBlack -0.565349*** -0.8035***Age -0.062758*** -0.0582***Female -0.045036 -0.1686Married 0.032832 -0.1288Education attainment 0.123125*** 0.0944***Children -0.509338*** -0.2158**

n 49,920 7,426# leaving 6,511 761% leaving 13.04% 10.25%

* p < .10; ** p < .05; *** p < .01.a The 1981 Almanac data included a score for housing costs. In the 1997 data, thehousing costs score is substituted with a broader cost-of-living score that includeshousing costs. For the cost-of-living score, high values indicate affordability (i.e., lowcosts). Thus, although the signs of the estimates differ across the two studies, bothsuggest that people leave high-cost areas at a higher rate than low-cost areas.b The 1997 Placed Rate Almanac used “Job Outlook” instead of “Economics.”c In the earlier data, low crime scores are indicative of low crime rates. In the 1997data, the opposite is true. Thus, despite the differing signs, both studies suggest thatmigration propensities increase with increasing crime rates.


TABLE 2. FIFTY MSAS WITH THE LOWEST OUT-MIGRATION RATES DERIVED FROM THE

UPDATED, ORIGINAL HERZOG AND SCHLOTTMANN (1986) MODEL (METRO

AREAS IN ALPHABETICAL ORDER).

Original andupdated model

Original model only Updated model only

Albany-Schenectady, NY Asheville, NC Akron, OHAppleton-Oshkosh, WI Billings, MT Atlanta, GABuffalo-Niagara Falls, NY Binghamton, NY Barnstable-Yarmouth, MAChicago, IL Boston, MA Benton Harbor, MICincinnati, OH Brownsville-

Harlingen, TXBiloxi-Gulfport, MS

Cleveland-Lorain-Elyria,OH

Cedar Rapids, IA Denver, CO

Dallas, TX Eau Claire, WI Detroit, MIDuluth-Superior, MN-WI Elmira, NY Glens Falls, NYErie, PA Evansville, IN Hartford, CTFort Wayne, IN Grand Forks, ND Houma, LAFort Worth, TX Green Bay, WI Houston, TXGrand Rapids-Muskegon,

MIGreensboro–Winston-

Salem, NCIndianapolis, IN

Milwaukee-Waukesha, WI Johnstown, PA Jamestown, NYMinneapolis-St. Paul, MN Kenosha, WI Kalamazoo-Battle Creek, MINassau-Suffolk, NY La Crosse, WI Kansas City, MOPittsburgh, PA Lansing-East

Lansing, MIKnoxville, TN

Rochester, NY Madison, WI Las Vegas, NVSt. Louis, MO McAllen-Pharr-

Edinburg, TXMonmouth-Ocean, NJ

Syracuse, NY New Bedford, MA New Orleans, LAOdessa, TX Norfolk-Virginia Beach, VAParkersburg-Marietta,

WV-OHOmaha, NE

Pittsfield, MA Orlando, FLRacine, WI Philadelphia, PA-NJScranton–Wilkes-

Barre, PAProvo-Orem, UT

Seattle-Everett, WA Richmond-Petersburg, VASioux Falls, SD Salt Lake City-Ogden, UTSt. Cloud, MN Sheboygan, WI


out-migration rates derived from the original and from the updated model. In the late1970s, metro areas in the Middle Atlantic and Midwest had the lowest out-migration rates.Twenty years later, the set of fifty metro areas with low out-migration rates overlaps quitea bit with the set identified for the late 1970s, with eighteen of the fifty metro areasappearing in both lists.1 However, using the 2000 data, we find that many metro areaslocated in the South and Mountain West have entered the top fifty list, including Atlanta,Denver, Las Vegas, Orlando, Salt Lake City, and Tampa-St. Petersburg. The significantparameter estimates of the updated model found in Table 1 indicate that this can be largelyattributed to affordable cost of living, low crime rates, and good access to educational andrecreation opportunities. Whereas much of the focus has been on in-migration to suchareas, these results suggest that part of their population growth can be attributed to theirretention ability.

While the results shown in Tables 1 and 2 are in general quite insightful, the modelscannot address possible differential valuation of amenities nor situate out-migration pro-pensities within metro areas’ demographic profiles. We next extended our analysis toaddress these shortcomings.

Research DesignModel. A discrete-choice binary logit model is used to analyze the effects of ameni-

ties on the decision to leave a metropolitan area. As in the Herzog and Schlottmann (1986)model, the dichotomous choice being analyzed is whether or not a person migrated out ofthe metropolitan area and the exogenous variables include the 1997 Places Rated Almanacscores for the place of origin. However, our modeling framework differs in three importantways from that of Herzog and Schlottmann. First, we include control variables that addresscontextual differences between metropolitan areas such as population size, density, and netmigration. Second, our model is estimated separately for six different life-course groups ofthe college-educated population. Finally, we lifted the birthplace restriction. That is, tworegimes were specified that represent, respectively, persons born within the same state astheir 1995 MSA of residence and persons born in a different state. The estimations allowfor possible differences in the valuation of amenities by place of birth.

TABLE 2. (CONTINUED)

Original andupdated model

Original model only Updated model only

Steubenville-Weirton,OH-WV

Tampa-St. Petersburg, FL

Terre Haute, IN Tulsa, OKUtica-Rome, NY Washington, DC-MD-VAWheeling, WV Youngstown-Warren, OHWichita, KS


Sample. The 2000 Census 5 percent PUMS data include information on house-holders’ age, marital status, educational attainment, employment status, presence ofchildren, and residential locations in 1995 and 2000. This information is used to define thebroad population at risk of migrating, namely all householders who, in 2000, had at leasta bachelor’s degree, lived anywhere in the United States (metro or nonmetro area) and who,in 1995, resided in a metropolitan area.2 We followed the lead of Herzog and Schlottmannby excluding householders who are 1) under 65 and unemployed, 2) over 64 and employed,3) not born in the United States, 4) serving in the military in 2000, and 5) not living in ahousing unit in 2000.

Sample segmentation. The sample of college-educated persons is segmented into sixgroups that reflect distinctive stages of the life course.3 The groups are defined by maritalstatus, age, presence of children, and employment status (Table 3).4,5 Recent graduates areemployed, between the ages of twenty-four and twenty-eight in 2000, have never beenmarried, and have no children. Their moves between 1995 and 2000 represent the college-to-work migration. The group is split by gender to allow for gender gaps in migrationbehavior. The group of young marrieds consists of married childless householders betweenthe ages twenty-nine and thirty-three. There is no gender distinction because people in thisgroup are married and presumed to move jointly, where either spouse could have been listedas the householder. The numerically largest group, the settled-down middle-agers, consistsof people between ages thirty-four and fifty-four. They have at least one child and aremarried. The about-to-retire empty-nesters group includes persons between the ages fifty-five and sixty-four in 2000. They are married but do not have any children living with them;it is unknown if earlier in life their households included children. Finally, the group ofretired empty-nesters consists of people between the ages sixty-five and eighty-five, whoare not in the labor force, and are married without children living at home.

Variable definitions and hypothesized effects. The list of variables, their sources, anddefinitions are summarized in Table 4. The dependent variable, derived from 2000 PUMSdata on residential location in 1995 and 2000, takes the value of 1 (“mover”) if the 1995 MSAof residence is different from the 2000 residence and the value of 0 (“stayer”) otherwise.6

The location-specific scores from the 1997 Places Rated Almanac are the key explana-tory variables. MSAs are rated on nine factors believed to influence the quality of a place:cost of living, job outlook, climate, recreation, the arts, health care, education, trans-portation, and crime (Savageau and Loftus 1997). Despite its primarily nonscientificpurpose, the Places Rated Almanac has been widely used in QOL research (Florida 2002a;Rappaport 2003). Moreover, following some criticism (see Landis and Sawicki 1988)7

of the earlier editions, the Places Rated Almanac methodology has undergone somemodification over the years.

The cost-of-living score in the Almanac is determined by nine categories that accountfor most expenditures of a “typical” four-person household. It is probably the most con-troversial and trickiest to interpret because of its broad range of inputs. In particular, thehousing-cost component, which accounts for 25 percent of the score, is problematic.Previous work has shown that places with comparative production advantages not


TA

BLE

3.S

AM

PLE

SE

GM

EN

TA

TIO

N.

Sha

red

char

acte

ristic

s

Edu

catio

nala

ttain

men

tA

tle

ast

aba

chel

or’s

degr

ee

Res

iden

ce19

95R

esid

edin

anM

SA

in19

95an

dliv

edan

ywhe

rein

the

Uni

ted

Sta

tes

in20

00

Dis

tingu

ishi

ngch

arac

teris

tics

Rec

ent

grad

uate

sY

oung

mar

rieds

Set

tled-

dow

nm

iddl

e-ag

ers

Abo

ut-t

o-re

tire

empt

y-ne

ster

sR

etire

dem

pty-

nest

ers

Rec

ent

alum

niR

ecen

tal

umna

e

Sex

Mal

eF

emal

eB

oth

Age

24–2

829

–33

34–5

455

–64

65–8

5M

arita

lsta

tus

Sin

gle

Mar

ried

Pre

senc

eof

child

ren

No

Yes

No

Em

ploy

edY

esN

o


TABLE 4. VARIABLE DEFINITIONS AND SOURCES.

Variable Definition Source

Dependent variableMover = 1 if 1995 MSA of residence differs from

2000 residence= 0 otherwise

2000 5% PUMS

Exogenous variablesCost of living Components (weights): mortgage (25.3%);

transportation (21.3%); food (19.8%);healthcare (9.3%); utilities (8.9%); otherexpenditures (6.9%); income tax (4.1%);property tax (3.2%); sales tax (1.2%)

1997 PlacesRated Almanac

Job outlook Components (weights): increase in numberof new jobs (74%); percentage increaseof new jobs (26%)

Transportation Components (weights): connectivity (60%);commute (30%); centrality (10%)

Education Components (weights): two-year institutionsenrollment (9.5%); bachelor’s degreeinstitutions enrollment (19%);comprehensive institutions enrollment(28.6%); doctoral degree institutionsenrollment (42.9%)

Health Components (weights unavailable):general/family practitioners per capita;medical specialists per capita; surgicalspecialists per capita; accredited generalhospital beds per capita; medical studentenrollment per capita

Crime Components (weights): violent crime rate(90.9%); property crime rate (9.1%)

Arts Components (weights unavailable):bigness—sheer number of culturalactivities; reading popularity; museumpopularity

Recreation Components (weights): bigness—sheernumber of recreation activities (60%);recreation land (22%); golf, movies, goodfood per capita (18%)


only have higher housing costs but also higher wages (Roback 1982). Moreover,amenities—including the man-made variety—are capitalized in housing values and rents(Bloomquist, Berger, and Hoehn 1988; Welch, Carruthers, and Waldorf 2007).

The Almanac’s job outlook factor describes the short-term expectation for job growthbetween 1997 and 2000. It is simply a projection of aggregate job growth and does notaccount for the MSA’s industry mix of employment. The climate score assigns places withmild, sunny, and stable conditions better ratings than places with extreme, cloudy, or unstableconditions. The recreation score assesses the availability of popular recreational activities,popular leisure activities, and the quality of restaurants. The arts, health care, and educationdimensions capture the incidence of cultural amenities, the availability of health profession-als, and the existence of local higher-education opportunities, respectively. However, thescores do not account for any inter-metropolitan differences in quality, expense, or accessi-bility. While the education score certainly reflects the variety of higher education options, itdoes not adequately assess the quality of local schools or colleges. Moreover, the proprietarynature of the data means that the weighting methodologies of the Almanac are not alwaystransparent to the reader. The transportation score describes the efficiency of local transpor-tation infrastructure, the availability of public transportation, and the degree of connectivityto various national transportation grids.The crime score describes the prevalence of propertyand violent crime. Because the impact on victims differs across these two types of crime,violent crime is weighted ten times more heavily than property crime.

For all nine factors, the scores range on a continuous scale from zero to one hundred andare standardized such that the 50th percentile point is the average score of all MSAs(Savageau and Loftus 1997). Low scores signal strong disamenities: that is, lack ofopportunities in recreation, the arts, education, and health care; poor transportationservices; poor climate; high crime rates; and high cost of living. Table 5 summarizes the

TABLE 4. (CONTINUED)

Variable Definition Source

Climate Components (weights): mildness (72%);brightness (16%); stability (12%)

Populationsize

MSA’s population size in 2000 U.S. Census2000

Density MSA’s population per square mile in 2000Net migration Net migration rate between 1995 and 2000Human capital

growthPercentage change of college educated

persons between 1990 and 2000U.S. Census

1990 and 2000Mile50 Number of metropolitan areas within a

50-mile radius of the metro area ofresidence

OMB, 1999 MSABoundaries


TA

BLE

5.H

YP

OT

HE

SIZ

ED

EF

FE

CT

SO

NO

UT-M

IGR

AT

ION

PR

OP

EN

SIT

IES.

Var

iabl

esR

ecen

tal

umni

You

ngm

arrie

dsS

ettle

d-do

wn

mid

dle-

ager

sA

bout

-to-

retir

eem

pty-

nest

ers

Ret

ired

empt

y-ne

ster

s

Con

stan

t(m

igra

tion

prop

ensi

ty)

Ver

yhi

ghH

igh

Low

Ver

ylo

wLo

wC

ost

ofliv

ing

--

-Jo

bou

tlook

--

-T

rans

port

atio

nE

duca

tion

Hea

lth-

Crim

e-

-A

rts

--

Rec

reat

ion

--

-C

limat

e-

Pop

ulat

ion

size

--

++/

-+/

-D

ensi

ty-

-+

--

Mile

50+/

-+/

-+/

-+/

-+/

-N

etm

igra

tion

--

--

-H

uman

capi

talg

row

th-

--

--


hypothesized effects of the QOL variables on migration propensities for each of the sixdemographic segments. A negative effect on migration indicates that the absence or weak-ness of an amenity serves as a push factor of migration.

For recent graduates, finding a good, permanent job is hypothesized to be the mostimportant factor. Because these persons are at the beginning of their careers and may beburdened by student loans, high cost of living is also hypothesized to be a push factor.Moreover, they are expected to highly value recreation and the arts. The young marriedshave already started a career, may enjoy a sizeable discretionary income, and can migratefor career advancement without being tied down by children. Just like the recent graduates,we expect them to be more likely to leave MSAs with low scores on the job outlook,recreation, and arts factors, yet be less concerned about cost of living. Settled-downmiddle-agers with children are expected to highly value family-friendly environments.Safety concerns among this group are hypothesized to raise out-migration from MSAs withlow crime scores (i.e., high crime rates). To offset the high cost of raising children, theyare expected to leave expensive MSAs. For the about-to-retire empty-nesters, no a prioriexpectations are specified other than that their migration propensity is the lowest among thesix groups. Many in this group are likely to retire soon, and relocation decisions are likelyto be postponed until retirement. Finally, the retired empty-nesters do not have mobilityrestrictions associated with employment or children. We expect that high cost of living,high crime rates, and adverse climates are significant push factors for this group. Inaddition, we hypothesize that this older population group values good health care and thushas high out-migration rates in those MSAs with poor scores on the health care factor.

The model also includes contextual factors that situate out-migration decisions withinthe overall spatial-demographic context. Such controls are necessary because out- andin-migration rates tend to be positively correlated, at least in the aggregate,8 and out-migration rates are often highly correlated with places’ overall population growth andeconomic prosperity (Beale 1969; Gleave and Cordey-Hayes 1977; Lowry 1966; Plane,Rogerson, and Rosen 1984). Thus, without consideration of the spatial-demographiccontext, estimated out-migration propensities cannot be used to make statements aboutpopulation change and can certainly not be interpreted as an indicator for a metro area’sdesirability.

Five variables are used to describe the spatial-demographic context. Population size anddensity control for the wealth of opportunities within the metropolitan area. Residents ofsmall and low-density metro areas may not easily find alternative employment or housinglocally. They, therefore, may be more inclined to leave a metropolitan area. Both variablesare thus hypothesized to have a negative effect on out-migration propensities of the younggroups interested in starting or advancing their careers.

The distance variable Mile50—indicating the number of metropolitan areas within afifty-mile radius—allows us to differentiate between metro areas that are part of a largerurban agglomeration, for example, those along the East Coast, and metro areas that have alarge rural hinterland, for example, those in the Great Plains. We do not have a priorexpectation about the net direction of the effect. On the one hand, it may be that remotely


located metro areas record low out-migration rates because every inter-metropolitan moveimplies that it be a long-distance move. On the other hand, in metro areas that aresurrounded by other metro areas, out-migration rates may also be low because commutingmay serve as a substitute for migration.

The net migration variable controls whether a metropolitan area gains or loses throughmigration. It differentiates between, for example, Denver and Buffalo. In Denver, lowout-migration rates are part of an overall growing demographic setting whereas in Buffalothey are part of a stagnating demography. Finally, the variable human capital growthindicates changes in the share of the college-educated population. We have witnessedincreasing spatial concentrations of the college-educated population, epitomized in theemergence of knowledge-based regional economies like California’s Silicon Valley and theResearch Triangle in North Carolina. We expect to see these agglomeration effects mani-fested in an inverse relationship between out-migration of the college-educated populationand past growth of human capital.

Empirical ResultsGeneral observations. Table 6 summarizes the estimation results for the six life-

course groups. Overall, the models of the propensity to leave an MSA perform quite well,with the log-likelihood ratio being significant for all groups. The model performs best forthe Recent Alumnae, followed by their male counterparts.

Before discussing the group-specific valuation of QOL indicators, several observationscan be made. First, consistent with other findings in the literature (Gottlieb and Joseph2006; Hansen, Ban, and Huggins 2003), the migration propensity of persons living in theirbirth state is significantly smaller than the migration propensity of residents who were bornin a different state. This finding holds over all life-course groups. However, the behavioraldifferences are quite substantial, in general, declining by age from a high of more than 11percentage points for the recent graduates to a low of about 2.9 percentage points for theoldest.

Second, consistent with the literature on the age-dependency of migration (Pandit 1997;Rogers 1988), we find that migration propensities decline with increasing age up until thestage of retirement, and then the propensity reverses for the oldest group. However, itshould be noted that the observed differences in migration rates across the six life-coursegroups are also influenced by other group-defining characteristics, particularly maritalstatus and the presence of children. Moreover, the results also suggest that gender differ-ences persist among recent graduates with men being slightly more likely to migrate thanwomen.

Third, there is an almost universal tendency to stay in those metro areas that enjoy agrowing human capital stock. This is an important result. It suggests that the agglomerationof human capital is a self-propelling process that—once set in motion—continues bydampening out-migration propensities of the college-educated. It also suggests that stop-ping the brain drain in metropolitan areas with slow or even negative human capital growth


TA

BLE

6.P

AR

AM

ET

ER

ES

TIM

AT

ES

aA

ND

WA

LDS

TA

TIS

TIC

SO

FB

INA

RY

LOG

ITM

OD

ELS

.

Var

iabl

eR

ecen

tal

umni

Rec

ent

alum

nae

Livi

ngin

stat

eof

birt

hLi

ving

out

ofbi

rth

stat

eLi

ving

inst

ate

ofbi

rth

Livi

ngou

tof

birt

hst

ate

bW

ald

bW

ald

bW

ald

bW

ald

Con

stan

t1.

622

45.4

322.

446

57.7

711.

929

71.5

972.

110

47.1

78Q

ualit

y-of

-life

indi

cato

rsC

ost

ofliv

ing

-3.1

82E

-05

4.29

4-2

.133

E-0

51.

138

-5.8

14E

-05

16.2

817.

404E

-06

0.15

9Jo

bou

tlook

-5.3

69E

-06

0.27

9-4

.437

E-0

512

.251

-1.5

08E

-05

2.43

1-3

.901

E-0

511

.081

Tra

nspo

rtat

ion

-8.9

52E

-06

0.30

8-6

.528

E-0

60.

083

-3.6

54E

-05

5.75

8-3

.377

E-0

52.

431

Edu

catio

n2.

563E

-05

2.07

13.

762E

-05

2.62

43.

759E

-05

5.06

94.

475E

-05

4.02

5H

ealth

-3.2

01E

-05

7.14

9-2

.531

E-0

52.

462

1.30

6E-0

60.

013

-3.4

66E

-05

5.50

4C

rime

4.33

6E-0

518

.525

4.75

1E-0

510

.617

4.65

0E-0

523

.716

3.92

2E-0

58.

346

Art

s-7

.757

E-0

513

.743

-8.0

61E

-05

7.21

1-9

.849

E-0

523

.692

-4.9

67E

-05

3.19

4R

ecre

atio

n-6

.085

E-0

519

.908

-3.4

39E

-05

3.30

6-4

.881

E-0

513

.770

-4.2

60E

-05

5.63

2C

limat

e2.

135E

-05

3.65

0-5

.835

E-0

515

.267

-1.0

13E

-06

0.00

9- 1

.738

E-0

60.

015

Con

trol

varia

bles

Pop

ulat

ion

size

-7.9

89E

-08

24.4

66-9

.587

E-0

821

.951

-7.3

66E

-08

23.6

98-9

.264

E-0

823

.073

Den

sity

2.30

2E-0

51.

223

-3.3

17E

-05

1.50

52.

302E

-06

0.01

4-2

.202

E-0

50.

871

Mile

500.

0059

473

1.00

50.

0088

807

1.19

3-0

.008

3845

2.14

10.

0216

078

8.26

1N

etm

igra

tion

0.00

3389

510

.935

0.00

1095

20.

713

0.00

3031

89.

466

0.00

1841

91.

961

Hum

anca

pita

lgr

owth

-1.2

8932

5924

.684

-0.8

2759

256.

573

-1.4

3686

3131

.654

-1.4

3975

2721

.617


TA

BLE

6.(C

ON

TIN

UE

D)

Var

iabl

eR

ecen

tal

umni

Rec

ent

alum

nae

Livi

ngin

stat

eof

birt

hLi

ving

out

ofbi

rth

stat

eLi

ving

inst

ate

ofbi

rth

Livi

ngou

tof

birt

hst

ate

bW

ald

bW

ald

bW

ald

bW

ald

N16

,146

17,7

07n

bypl

ace

ofbi

rth

9,76

86,

378

10,4

877,

220

%le

avin

gby

plac

eof

birt

h51

.33

63.7

249

.61

61.1

5

Chi

-squ

ared

1,62

3.21

21,

667.

118

d.f.

2929

p-va

lue

0.00

00.

000

Var

iabl

eY

oung

mar

ried

Set

tled-

dow

nm

iddl

e-ag

ers

Livi

ngin

stat

eof

birt

hLi

ving

out

ofbi

rth

stat

eLi

ving

inst

ate

ofbi

rth

Livi

ngou

tof

birt

hst

ate

bW

ald

bW

ald

bW

ald

bW

ald

Con

stan

t0.

593

5.71

41.

499

32.5

25-1

.524

46.1

48-0

.433

6.15

7Q

ualit

y-of

-life

indi

cato

rsC

ost

ofliv

ing

-1.2

64E

-05

0.62

23.

310E

-06

0.04

29.

419E

-06

0.41

21.

443E

-05

1.81

1


TA

BLE

6.(C

ON

TIN

UE

D)

Var

iabl

eY

oung

mar

ried

Set

tled-

dow

nm

iddl

e-ag

ers

Livi

ngin

stat

eof

birt

hLi

ving

out

ofbi

rth

stat

eLi

ving

inst

ate

ofbi

rth

Livi

ngou

tof

birt

hst

ate

bW

ald

bW

ald

bW

ald

bW

ald

Job

outlo

ok-3

.807

E-0

514

.341

-3.4

79E

-05

11.5

79-1

.981

E-0

54.

333

2.12

9E-0

60.

075

Tra

nspo

rtat

ion

-3.2

13E

-06

0.03

96.

202E

-06

0.11

43.

227E

-05

4.36

4-4

.899

E-0

515

.225

Edu

catio

n3.

697E

-05

4.21

72.

781E

-05

2.36

61.

062E

-05

0.41

1-7

.143

E-0

60.

403

Hea

lth2.

103E

-05

2.78

9-2

.116

E-0

52.

831

2.47

2E-0

54.

317

3.33

0E-0

60.

151

Crim

e-2

.990

E-0

58.

576

8.76

3E-0

70.

006

-3.3

11E

-05

12.2

53-4

.605

E-0

533

.979

Art

s-5

.025

E-0

55.

115

-8.1

62E

-05

11.1

33-2

.719

E-0

51.

774

1.14

8E-0

50.

533

Rec

reat

ion

-8.5

66E

-05

32.9

30-3

.257

E-0

54.

049

-8.6

29E

-05

38.8

09-2

.284

E-0

54.

156

Clim

ate

3.31

1E-0

58.

682

-2.3

92E

-05

4.23

83.

217E

-05

9.72

7-4

.924

E-0

535

.576

Con

trol

varia

bles

Pop

ulat

ion

size

-8.7

10E

-08

30.9

41-3

.917

E-0

86.

009

-7.9

04E

-09

0.29

4-4

.075

E-0

810

.991

Den

sity

8.68

6E-0

523

.372

2.09

5E-0

51.

126

0.00

0109

843

.499

6.31

5E-0

512

.557

Mile

50-0

.011

6754

3.67

1-0

.007

1280

1.18

6-0

.031

9982

30.2

12-0

.020

2243

19.4

79N

etm

igra

tion

0.00

2779

46.

459

0.00

0135

40.

016

0.00

0167

00.

027

-0.0

0351

5620

.796


TA

BLE

6.(C

ON

TIN

UE

D)

Var

iabl

eY

oung

mar

ried

Set

tled-

dow

nm

iddl

e-ag

ers

Livi

ngin

stat

eof

birt

hLi

ving

out

ofbi

rth

stat

eLi

ving

inst

ate

ofbi

rth

Livi

ngou

tof

birt

hst

ate

bW

ald

bW

ald

bW

ald

bW

ald

Hum

anca

pita

lgr

owth

-0.7

6825

067.

227

-0.7

7287

198.

377

0.00

3795

52.

16E

-04

0.48

3593

96.

423

N19

,056

50,0

00n

bypl

ace

ofbi

rth

10,3

028,

754

24,8

1725

,183

%le

avin

gby

plac

eof

birt

h

31.7

246

.63

10.8

421

.64

Chi

-squ

ared

1,14

3.13

21,

543.

475

d.f.

2929

p-va

lue

0.00

00.

000


TA

BLE

6.(C

ON

TIN

UE

D)

Var

iabl

eA

bout

-to-

retir

eem

pty-

nest

ers

Ret

ired

empt

y-ne

ster

s

Livi

ngin

stat

eof

birt

hLi

ving

out

ofbi

rth

stat

eLi

ving

inst

ate

ofbi

rth

Livi

ngou

tof

birt

hst

ate

bW

ald

bW

ald

bW

ald

bW

ald

Con

stan

t-2

.518

72.3

93-1

.625

68.5

97-2

.109

49.9

64-1

.738

72.9

93Q

ualit

y-of

-life

indi

cato

rsC

ost

ofliv

ing

2.19

3E-0

51.

322

2.78

7E-0

60.

053

-6.5

43E

-05

11.4

32-4

.310

E-0

511

.901

Job

outlo

ok2.

267E

-05

2.99

54.

301E

-07

0.00

32.

457E

-05

3.83

25.

592E

-06

0.41

5T

rans

port

atio

n-9

.727

E-0

60.

248

1.13

9E-0

50.

646

-7.2

94E

-07

0.00

13.

490E

-06

0.05

6E

duca

tion

8.35

0E-0

60.

146

-1.5

64E

-05

1.57

5-8

.359

E-0

60.

147

2.72

5E-0

54.

727

Hea

lth8.

911E

-06

0.35

7-2

.894

E-0

60.

089

5.59

6E-0

512

.516

8.26

9E-0

60.

602

Crim

e-2

.820

E-0

55.

092

-1.9

44E

-05

4.85

6-3

.540

E-0

60.

088

-1.6

13E

-05

3.21

9A

rts

3.24

4E-0

51.

515

9.66

1E-0

60.

308

-3.4

51E

-05

1.67

6-1

.044

E-0

50.

333

Rec

reat

ion

-6.4

60E

-05

12.5

42-1

.864

E-0

52.

143

2.74

0E-0

60.

022

-4.5

77E

-05

11.8

44C

limat

e9.

349E

-08

4.5E

-05

-3.4

94E

-05

14.3

88-5

.513

E-0

516

.247

-7.0

06E

-05

54.9

12C

ontr

olva

riabl

esP

opul

atio

nsi

ze-4

.128

E-0

83.

780

-3.5

93E

-08

7.26

0-5

.161

E-0

86.

580

-2.8

72E

-08

4.53

9D

ensi

ty3.

301E

-05

1.63

38.

549E

-06

0.14

96.

559E

-05

8.69

63.

366E

-05

2.31

2M

ile50

-0.0

0567

770.

563

-0.0

1143

465.

074

0.00

4110

40.

342

0.00

0686

30.

018

Net

mig

ratio

n-0

.003

0171

5.11

9-0

.003

8744

22.4

25-0

.003

0945

4.94

6-0

.006

0387

58.0

25H

uman

capi

tal

grow

th0.

3268

792

0.89

20.

7066

388

11.4

710.

0639

323

0.03

11.

3270

529

42.8

06


TA

BLE

6.(C

ON

TIN

UE

D)

Var

iabl

eA

bout

-to-

retir

eem

pty-

nest

ers

Ret

ired

empt

y-ne

ster

s

Livi

ngin

stat

eof

birt

hLi

ving

out

ofbi

rth

stat

eLi

ving

inst

ate

ofbi

rth

Livi

ngou

tof

birt

hst

ate

bW

ald

bW

ald

bW

ald

bW

ald

N50

,000

50,0

00n

bypl

ace

ofbi

rth

21,3

4628

,654

19,8

6530

,135

%le

avin

gby

plac

eof

birt

h7.

1912

.93

8.78

11.6

9

Chi

-squ

ared

568.

481

462.

719

d.f.

2929

p-va

lue

0.00

00.

000

aE

stim

ates

inbo

ldar

esi

gnifi

cant

atth

e0.

05le

vel(

two-

taile

dte

st).


promises to be difficult because of the especially high out-migration rates exhibited by thecollege-educated.

Fourth, the groups differ with respect to city-size and density preferences. For personsborn out-of-state, the migration propensity decreases with increasing population size. Forpersons born in the state of their 1995 residence, the big-city preference only holds for theyounger age groups and the very old. With respect to density, young married couples,settled-down middle-agers, and those about-to-retire are more likely to leave from high-density metro areas than from low-density metro areas. The migration behavior of theyoungest and the oldest groups appears to be indifferent to density.

Fifth, we also find age differences with respect to net migration and proximity to othermetro areas. As overall net migration rates increase, we find that the out-migration pro-pensities increase for the younger but decrease for the older cohorts. The presence of othermetro areas in close proximity has no effect on the migration propensities of the youngestand the oldest. Interestingly though, it dampens the migration propensities of the middle-aged, who may substitute commuting for outright relocation.

Group-specific valuations of QOL indicators. Care should be taken in evaluating therole of amenities in the migration choices of college-leavers because their migrationbehavior is qualitatively different than the residential location choices made later in the lifecourse. Students are generally considered “special populations” in that their moves areoften viewed as fixed-period sojourns in a place. Out-migration from a college town doesnot necessarily imply leaving an “undesirable” place but instead choosing a workplacedestination.

Given the mentioned caveat, our results suggest that recent graduates prefer to remainin affordable places. Their out-migration propensities generally increase with increasingcost of living. Interestingly, only for those born out-of-state does a poor job-outlooktrigger out-migration. However, attributes appearing to be important to all young cohortsare recreational opportunities and a rich cultural environment. Metropolitan areas withlow scores on the recreation and arts factors have higher out-migration rates than met-ropolitan areas that are well endowed with recreational opportunities and cultural activi-ties. This is an important result that fits nicely into the debate about the economicdevelopment role of the “creative class” (Florida 2002a,b) and its preference for bohe-mian environments.

We find that recent graduates’ propensity to leave a metropolitan area increases withdecreasing crime rates. This contrasts with results obtained from a recent micro-level studythat finds that high crime rates increase the hazard of migrating to another state (Huffmanand Feridhanusetyawan 2007). However, our results are not implausible and do not implythat young graduates have a dislike for safe environments. Instead, we suspect that thislife-course group has the financial means to locate in safe enclaves within metropolitanareas. Although there is evidence that high crime rates induce households to relocate(Cullen and Levitt 1999; Morenoff and Sampson 1997), Dugan (1999) suggests that theimpact of crime rates on population turnover might in fact be quite localized, perhaps notextending beyond a one-mile radius around the crime scene.


The results for the impact of transportation, education, health care, and climate indica-tors on the migration behavior of the recent graduates are mixed. Good transportationinfrastructure lowers the migration propensities of young women residing in their birthstate in 1995, but it is not a decisive criterion for male recent graduates. Good educationalopportunities also do not play a role for young men and even increase the migrationpropensity of young women. This is not surprising given that this group has just completedits education and has now entered the labor force. Good health care infrastructure lowersthe migration propensity or is insignificant. Finally, a good climate is—for the mostpart—irrelevant for the migration behavior of this group.

The results for young and childless married individuals show some similarities withthose obtained for recent graduates. Young married couples also appear to strongly valuerich recreation and arts amenities. However, we find them indifferent toward cost-of-livingdifferentials. Presumably, because they are college-educated, have no children, and are(most likely working) spouses, they have sufficient disposable income to enjoy suchamenities without worrying overly about high cost of living. They are, however, responsiveto poor job prospects. The health and transportation infrastructures do not influence theirmigration behavior and rich educational opportunities do not play a role given that thisgroup has completed their schooling. Those who were born out-of-state tend to be likely toleave from metro areas with harsh climate conditions and individuals who were born intheir state of residence tend to have high out-migration rates from metropolitan areas withhigh crimes rates.

Aging and having children brings about drastic lifestyle changes. It is thus not surpris-ing that the results for the settled-down middle-agers differ markedly from those of theyounger and childless individuals. Not only do the settled-down middle-agers migratesubstantially less (fewer than 13 percent of this group left their metro area between 1995to 2000 compared to 30 percent or more for the younger groups) there is also a shift inpreferences toward safety. For this group, out-migration propensities increase significantlywith increasing crime rates. The results also suggest that settled-down middle-agers valuerich recreational opportunities but are indifferent toward opportunities of higher education,an artsy environment, and cost of living. For those born in the state of their 1995 residence,a poor job outlook is a reason to leave, and for those born in another state, a poor climateis a strong trigger for out-migration.

The about-to-retire empty-nesters have the lowest overall out-migration propensity.Migration for this group is thus a somewhat unpredictable event and we suspect that many ofthe moves that do take place are made in anticipation of the upcoming retirement.Aside froma concern about safety—migration propensities increase with increasing crime rates—QOLindicators do not systematically influence the migration behavior of this group.

For the retired empty-nesters, we can safely assume that the moves are motivated by thenew locational freedom associated with no longer being employed full-time. Our resultsshow that retirees have a strong preference for a mild climate: they have a higher propensityto leave metro areas with unfavorable climates than those with a mild climate. Moreover,retirees—who often have to live on a reduced income—do care about affordability. Their


propensity to leave a metropolitan area increases with increasing cost of living. Whereascost of living is highest in the largest metropolitan areas, note that our multivariate findingsshow that, when costs are taken into account, the college-educated retirees have a lowerpropensity to leave from bigger compared to smaller metropolitan areas. This finding is ofinterest given the significant role currently being played by elderly migration in redistrib-uting population downward within the U.S. metropolitan hierarchy (Plane, Henrie, andPerry 2005).

We cannot confirm the hypothesis that the migration propensities of the elderly areinfluenced by a metro area’s health care opportunities. This unexpected result may be anartifact of the health indicator, which addresses barriers to health care rather crudely, usingvarious physician-per-capita indicators. While such per capita indicators are commonlyused in the health care literature, when applied at the wider metropolitan scale they do notaccount for health-worker disparities that can be quite extensive, both in inner-city areasand in the outlying counties of large metropolitan areas.

Ranking metropolitan areas. Applying the estimated parameters to U.S. metropolitanareas allows us to derive group-specific rankings based on out-migration rates. For eachgroup, the metropolitan areas with the highest and lowest out-migration rates are listed inTable 7.

For the younger life-course groups, the set of metropolitan areas with the lowestout-migration rates includes the nation’s biggest metropolitan areas. Many of those metroareas have previously been identified as favorable destinations of the young, single, andcollege-educated adults (Franklin 2003). This big-city bias is also confirmed in studies ofthe destination choice of college-to-work migrants (Gottlieb and Joseph 2006) and thelocational choices of the so-called power couples (Costa and Kahn 2000) who are attractedto large urban areas with their abundance of managerial and professional jobs. On theother end of the scale, our estimates suggest very high rates of out-migration from smallmetro areas. Not surprisingly, these places include relatively small college towns suchas Lafayette, Indiana (Purdue University), Bryan-College Station, Texas (Texas A&MUniversity), and State College, Pennsylvania (Pennsylvania State University). While offer-ing excellent educational opportunities, these metro areas provide few employment oppor-tunities for their college graduates. In combination, these results confirm the agglomerationof the young college-educated in a select group of urban centers.

The list of metro areas with low out-migration rates for the settled-down middle-agersincludes a mixture of differently sized metropolitan areas, located in various regions of thecountry. A more distinct pattern is observable for the places with the highest out-migrationrates. They include small- and medium-sized metro areas located outside the westernUnited States. For the two oldest groups, the list of places with the lowest out-migrationrates has a very distinct Sunbelt bias, including many metro areas located in California andFlorida. Not surprisingly, places with a cold climate dominate the list of metropolitan areaswith the highest out-migration rates.

Table 8 summarizes the correlations between the estimated group-specific out-migrationrates. The coefficients are highly positive among the rates of the younger groups, thus


TA

BLE

7.R

AN

KIN

GO

FM

ET

RO

PO

LIT

AN

AR

EA

SB

AS

ED

ON

ES

TIM

AT

ED

OU

T-M

IGR

AT

ION

RA

TE

S.

Rec

ent

alum

nia

Rec

ent

alum

nib

Rec

ent

alum

naea

Rec

ent

alum

naeb

You

ngm

arrie

dsa

You

ngm

arrie

dsb

Ten

clo

wes

tou

t-m

igra

tion

Chi

cago

,IL

Atla

nta,

GA

Chi

cago

,IL

Las

Veg

as,

NV

Chi

cago

,IL

Atla

nta,

GA

New

Yor

k,N

YLo

sA

ngel

es,

CA

Atla

nta,

GA

Atla

nta,

GA

Min

neap

olis

,M

NLa

sV

egas

,N

V

Los

Ang

eles

,C

AC

hica

go,

ILN

ewY

ork,

NY

Pho

enix

,A

ZW

ashi

ngto

n,D

C–M

D-V

AP

hoen

ix,

AZ

Atla

nta,

GA

Pho

enix

,A

ZLo

sA

ngel

es,

CA

Chi

cago

,IL

Det

roit,

MI

Chi

cago

,IL

Det

roit,

MI

Las

Veg

as,

NV

Las

Veg

as,

NV

Ft.

Laud

erda

le,

FL

Sal

tLa

keC

ity,

UT

Sea

ttle,

WA

Dal

las,

TX

Riv

ersi

de,

CA

Hou

ston

,T

XH

oust

on,

TX

Phi

lade

lphi

a,P

A-N

JH

oust

on,

TX

Mia

mi,

FL

Hou

ston

,T

XD

etro

it,M

ID

alla

s,T

XG

rand

Rap

ids,

MI

Was

hing

ton,

DC

-MD

-VA

Hou

ston

,T

XD

alla

s,T

XC

harlo

tte,

NC

Cha

rlotte

,N

CA

tlant

a,G

AD

alla

s,T

XW

ashi

ngto

n,D

C-M

D-V

AN

ewY

ork,

NY

Mia

mi,

FL

Det

roit,

MI

Las

Veg

as,

NV

Ft.

Laud

erda

le,

FL

Pho

enix

,A

ZT

ampa

,F

LP

hoen

ix,

AZ

Tam

pa,

FL

Hou

ston

,T

XO

rland

o,F

LT

enc

high

est

out-

mig

ratio

nS

tate

Col

lege

,P

AS

haro

n,P

AS

tate

Col

lege

,P

AD

anvi

lle,

VA

Jers

eyC

ity,

NJ

Sum

ter,

SC

Mer

ced,

CA

Mun

cie,

INC

hico

-Par

adis

e,C

AS

haro

n,P

AS

umte

r,S

CS

haro

n,P

A


TA

BLE

7.(C

ON

TIN

UE

D)

Rec

ent

alum

nia

Rec

ent

alum

nib

Rec

ent

alum

naea

Rec

ent

alum

naeb

You

ngm

arrie

dsa

You

ngm

arrie

dsb

Jack

sonv

ille,

NC

Utic

a-R

ome,

NY

Man

ches

ter,

NH

Sum

ter,

SC

Ath

ens,

GA

Mun

cie,

IN

Dan

ville

,V

AS

tate

Col

lege

,P

AM

erce

d,C

AS

tate

Col

lege

,P

ALu

bboc

k,T

XD

anvi

lle,

VA

Bry

an-C

olle

geS

tatio

n,T

XD

anvi

lle,

VA

Sha

ron,

PA

Jack

sonv

ille,

NC

Tyl

er,

TX

Ode

ssa,

TX

Yol

o,C

AT

erre

Hau

te,

INH

attie

sbur

g,M

ST

erre

Hau

te,

INB

ryan

-Col

lege

Sta

tion,

TX

Kan

kake

e,IL

Chi

co,

CA

Bin

gham

ton,

NY

San

Luis

Obi

spo,

CA

Bin

gham

ton,

NY

Yol

o,C

AG

olds

boro

,N

C

Sha

ron,

PA

Lafa

yette

,IN

Jack

sonv

ille,

NC

Lafa

yette

,IN

Abi

lene

,T

XS

tate

Col

lege

,P

ALa

faye

tte,

INW

illia

msp

ort,

PA

Bry

an-C

olle

geS

tatio

n,T

XM

erce

d,C

AW

aco,

TX

Utic

a-R

ome,

NY

Sum

ter,

SC

John

stow

n,P

AY

olo,

CA

Mun

cie,

INT

usca

loos

a,A

LA

bile

ne,

TX

Set

tled-

dow

nm

iddl

e-ag

ersa

Set

tled-

dow

nm

iddl

e-ag

ersb

Abo

ut-t

o-re

tire

empt

y-ne

ster

saA

bout

-to-

retir

eem

pty-

nest

ersb

Ret

ired

empt

y-ne

ster

saR

etire

dem

pty-

nest

ersb

Ten

clo

wes

tou

t-m

igra

tion

New

burg

h,N

YP

hila

delp

hia,

PA

-NJ

Las

Veg

as,

NV

Pun

taG

orda

,F

LO

cala

,F

LP

unta

Gor

da,

FL


TA

BLE

7.(C

ON

TIN

UE

D)

Set

tled-

dow

nm

iddl

e-ag

ersa

Set

tled-

dow

nm

iddl

e-ag

ersb

Abo

ut-t

o-re

tire

empt

y-ne

ster

saA

bout

-to-

retir

eem

pty-

nest

ersb

Ret

ired

empt

y-ne

ster

saR

etire

dem

pty-

nest

ersb

Gra

ndR

apid

s,M

IM

onm

outh

,N

JP

unta

Gor

da,

FL

San

Luis

Obi

spo,

CA

Pun

taG

orda

,F

LB

ryan

-Col

lege

Sta

tion,

TX

Mon

mou

th,

NJ

Pun

taG

orda

,F

LN

aple

s,F

LS

alin

as,

CA

Ft.

Pie

rce,

FL

Bilo

xi,

MS

Nas

sau,

NY

Bel

lingh

am,

WA

Ft.

Mye

rs,

FL

Sta

teC

olle

ge,

PA

Bry

an-C

olle

geS

tatio

n,T

XO

cala

,F

L

Har

tford

,C

TW

ashi

ngto

n,D

C–M

D-V

A

Bar

nsta

ble,

MA

Riv

ersi

de,

CA

Bra

zoria

,T

XF

t.P

ierc

e,F

L

App

leto

n,W

IP

hoen

ix-M

esa,

AZ

Pho

enix

,A

ZP

hila

delp

hia,

PA

-NJ

Las

Veg

as,

NV

San

Luis

Obi

spo,

CA

Kal

amaz

oo,

MI

Wilm

ingt

on,

DE

Los

Ang

eles

,C

ALo

sA

ngel

es,

CA

Dan

ville

,V

AJa

ckso

nvill

e,N

CD

ulut

h,M

NN

assa

u,N

YN

ewY

ork,

NY

Day

tona

Bea

ch,

FL

Hou

ma,

LAM

yrtle

Bea

ch,

SC

Hou

ma,

LAD

ayto

na.,

FL

Bel

lingh

am,

WA

Ft.

Mye

rs,

FL

Yum

a,A

ZD

ayto

naB

each

,F

LS

alin

as,

CA

Riv

ersi

de,

CA

Myr

tleB

each

,S

CS

aras

ota,

FL

Myr

tleB

each

,S

CF

t.M

yers

,F

L

Ten

chi

ghes

tou

t-m

igra

tion

Jers

eyC

ity,

NJ

Jers

eyC

ity,

NJ

Fay

ette

ville

,N

CLa

redo

,T

XJe

rsey

City

,N

JJe

rsey

City

,N

JLu

bboc

k,T

XLa

redo

,T

XLa

redo

,T

XS

ioux

City

,IA

New

Yor

k,N

YLa

redo

,T

X


TA

BLE

7.(C

ON

TIN

UE

D)

Set

tled-

dow

nm

iddl

e-ag

ersa

Set

tled-

dow

nm

iddl

e-ag

ersb

Abo

ut-t

o-re

tire

empt

y-ne

ster

saA

bout

-to-

retir

eem

pty-

nest

ersb

Ret

ired

empt

y-ne

ster

saR

etire

dem

pty-

nest

ersb

Ath

ens,

GA

Lim

a,O

HO

dess

a,T

XM

cAlle

n,T

XB

erge

n,N

JK

enos

ha,

WI

Jack

son,

TN

Jack

son,

TN

Jers

eyC

ity,

NJ

Fay

ette

ville

,N

CN

assa

u,N

YF

lint,

MI

New

Yor

k,N

YS

ioux

City

,IA

Alb

any,

GA

Anc

hora

ge,

AK

Milw

auke

e,W

IG

reen

Bay

,W

I

Top

eka,

KS

Mon

roe,

LAT

opek

a,K

SJe

rsey

City

,N

JA

ncho

rage

,A

KA

ncho

rage

,A

KA

bile

ne,

TX

Fay

ette

ville

,N

CE

lPas

o,T

XF

lint,

MI

Sta

mfo

rd,

CT

Tre

nton

,N

J

Ode

ssa,

TX

McA

llen,

TX

Sio

uxC

ity,

IAS

prin

gfiel

d,IL

Chi

cago

,IL

Ced

arR

apid

s,IA

Mon

roe,

LAA

ncho

rage

,A

KA

nnis

ton,

AL

Top

eka,

KS

Min

neap

olis

,M

NS

prin

gfiel

d,IL

Linc

oln,

NE

Alb

any,

GA

Sum

ter,

SC

Wau

sau,

WI

Det

roit,

MI

Col

orad

oS

prin

gs,

CO

aF

orpe

rson

sbo

rnin

the

stat

eof

thei

r19

95re

side

nce.

bF

orpe

rson

sbo

rnou

tsid

eth

est

ate

ofth

eir

1995

resi

denc

e.c

Aco

mpl

ete

list

ofth

egr

oup-

spec

ific

met

ro-r

anki

ngs

isav

aila

ble

from

the

auth

ors.


TA

BLE

8.C

OR

RE

LAT

ION

SA

MO

NG

GR

OU

P-S

PE

CIF

ICE

ST

IMA

TE

DO

UT-M

IGR

AT

ION

RA

TE

S.

Rec

ent

alum

nia

Rec

ent

alum

nib

Rec

ent

alum

naea

Rec

ent

alum

naeb

You

ngm

arrie

dsa

You

ngm

arrie

dsb

Set

tled-

dow

nm

iddl

e-ag

ersa

Set

tled-

dow

nm

iddl

e-ag

ersb

Abo

ut-t

o-re

tire

empt

y-ne

ster

sa

Abo

ut-t

o-re

tire

empt

y-ne

ster

sb

Ret

ired

empt

y-ne

ster

sa

Ret

ired

empt

y-ne

ster

sb

Rec

ent

alum

nia

1.00

0R

ecen

tal

umni

b0.

855

1.00

0R

ecen

tal

umna

ea0.

950

0.87

51.

000

Rec

ent

alum

naeb

0.87

30.

933

0.90

51.

000

You

ngm

arrie

dsa

0.77

60.

680

0.75

10.

729

1.00

0Y

oung

mar

rieds

b0.

881

0.94

10.

870

0.94

20.

825

1.00

0S

ettle

d-do

wn

mid

dle-

ager

sa0.

208

0.15

10.

137

0.15

80.

685

0.32

11.

000

Set

tled-

dow

nm

iddl

e-ag

ersb

0.25

60.

417

0.22

80.

366

0.53

70.

527

0.55

51.

000

Abo

ut-t

o-re

tire

empt

y-ne

ster

sa0.

467

0.50

10.

416

0.56

20.

712

0.64

50.

706

0.74

61.

000

Abo

ut-t

o-re

tire

empt

y-ne

ster

sb0.

072

0.36

40.

033

0.20

30.

260

0.36

30.

452

0.83

20.

673

1.00

0

Ret

ired

empt

y-ne

ster

sa-0

.493

-0.2

31-0

.300

-0.2

44-0

.404

-0.3

70-0

.079

-0.1

55-0

.102

0.10

41.

000

Ret

ired

empt

y-ne

ster

sb-0

.193

0.15

2-0

.119

0.02

0-0

.086

0.03

50.

201

0.33

40.

324

0.67

30.

650

1.00

0

aF

orpe

rson

sbo

rnin

the

stat

eof

thei

r19

95re

side

nce.

bF

orpe

rson

sbo

rnou

tsid

eth

est

ate

ofth

eir

1995

resi

denc

e.


suggesting strong similarities of their out-migration behaviors. The rates of the settled-downmiddle-agers are positively—but only weakly to moderately strong—correlated with thoseof the younger cohorts. Finally, the correlations between the estimated out-migration rates ofthe older groups and the younger groups are weak, and in some cases even significantlynegative. This clearly shows that the rankings of metropolitan areas are largely contingent onthe life-course stage. Thus, the concept of “rating” places in terms of their overall residentialdesirability—at least desirability in the strict sense that these metropolitan areas are best ableto retain their populations—makes very little sense as the retention power of urban placesvaries significantly across the different life-course groups.

Summary and ConclusionsThroughout history, migration-fueled population redistributions have characterized the

demographics of the United States. The settlement of the West, the migration of African-Americans to the North, the long-lasting rural population exodus of the twentieth century,and the recent migration of the elderly to the warmer climates of the South are just a fewexamples of such mass movements. Common to these diverse migration events is that theydifferentially involved segments of the population responding to specific attributes of theirplaces of origin or destination.

This paper set out to explore how the college-educated population—segmented intoselective demographic groups, from young adults to the elderly—differentially values QOLindicators of metropolitan areas. Among the noteworthy results, we find an almost universalpreference for not leaving metropolitan areas with substantial human capital growth.Moreover, ceteris paribus, young graduates show a preference for big metropolitan areas,whereas families with children prefer lower-density settings. High cost of living particu-larly encourages migration of the very young and the very old. Those who are young andchildless—whether married or unmarried—are very mobile in general, but have an espe-cially high propensity to leave metropolitan areas that offer insufficient recreational oppor-tunities and perform weakly in the arts. The large middle-aged life-course group leavesplaces with high crime rates and poor recreation amenities. Finally, retirees show a strongpreference for leaving any metropolitan area with a high cost of living and an unfavorableclimate. We thus conclude that—because the valuations of QOL indicators vary so greatlyacross the life course—an overall composite desirability index that rates the best and worstplaces is virtually meaningless. Instead, desirability indices pertaining to different life-course groups are more appropriate.

The study’s focus on out-migration of the college-educated is important in that retentionof an area’s most educated population positively influences its competitive position. In lightof shifting age distributions and an ever-growing share of the college-educated population,the results have important implications for the emergence of new migration patterns andthe concentration of human capital. Our findings suggest that—after controlling for otherspatial-demographic variables—human capital accumulation is a self-propelling processwith high out-migration propensities of the college-educated population in metro areaswith slow human capital growth. Thus, metropolitan areas that cannot, over a prolonged


time, build an extensive college-educated workforce risks falling behind in today’s knowl-edge economy. Our findings also provide important clues about tackling concerns about the(relative) loss of intellectual capital—concerns that have become ever so crucial in light ofnew threats to regions’ competitiveness in knowledge-based economies (Gottlieb 2001;Waldorf 2006). This research suggests that investments in cultural activities and recre-ational opportunities could possibly payoff by reducing the loss of the young college-educated residents.

From a methodological perspective, it is important to draw attention to two issues.First, micro data are now widely used in migration research. Because personal character-istics are included, these data sets are very powerful. However, there has been little workanalyzing how migration responses to QOL characteristics vary across stages of the lifecourse using micro data. If constraints are not imposed by the size of the data set, thisstudy has demonstrated the utility of allowing for variations in parameters through samplestratification.

Second, as in the Herzog and Schlottmann study, we focused solely on modelingout-migration. In other migration research, much attention has been devoted to QOLmeasures as determinants of the destination choice (Gottlieb and Joseph 2006). Therelationship between in- and out-migration behavior, and the relationship of both migrationrates to the attractiveness of places, has been the topic of earlier research (Plane, Rogerson,and Rosen 1984). Whereas in-migration has been found to be monotonically related toeconomic variables and other measures of desirability, a more U-shaped relationshipappears to characterize out-migration. Places with relatively healthy economies may notonly have high in-migration but also high out-migration. That is, people tend to move to,and from, positions of economic strength, resulting in positive correlations between in- andout-migration rates. Thus, care should be used in interpreting high rates of out-migration(as, e.g., in the case of college towns) as being indicative of low levels of residentialdesirability. Similarly, places with the lowest rates of out-migration may not be the mostdesirable as highly attractive places can be expected to have high rates of turnover. Ourmodels controlled for some such intervening factors of the spatial-demographic context.Nevertheless, future work needs to more explicitly connect the in- and out-migrationeffects of QOL measures to get a more accurate picture of the impact of place character-istics on the migration behavior of the well-educated as they pass through the life course.

NOTES1. The Dallas, Texas MSA and Fort Worth, Texas MSA were combined in the Herzog and Schlottmann

(1986) study but were analyzed separately in our updated model.

2. Because of mismatches between the geographic boundaries of 1999 MSAs and the Public-Use

Microdata Areas of Census 2000, 291 of the 331 MSAs defined by the Office of Management and

Budget in 1999 are used in this study.

3. The groups represent some possible manifestations of the different life paths individuals follow. By

no means do the six groups constitute an “all-inclusive” segmentation of the entire college-educated

population.


4. The models do not include race. Given the new race definitions used for Census 2000, identifying

race groups has become much more detailed and complex, limiting potential comparability to the

Herzog and Schlottmann (1986) results. Moreover, given that each group is homogenous with

respect to higher-education level, labor force status, marital status, age, and presence of children,

racial distinctions are of lesser importance than in the Herzog and Schlottmann study.

5. Although income has long been a key indicator influencing migration propensities, it is not used for

the segmentation or in the model specification. The income variable of the PUMS data is quite

problematic because it refers to the year 1999 rather than the time of the move. Moreover, it is

top-coded, and thus, cannot adequately reflect the income variation of the often well-paid college-

educated population. Interestingly, Herzog and Schlottmann (1986) also did not include income.

6. The long form of the U.S. Census does not ask specifically about migration, but asks about

residential location at the time of the census and five years prior. Using this locational information

to deduce migration behavior has some well-known problems, such as multiple moves during the

five-year period.

7. Landis and Sawicki (1988) argue that many types of quality of place issues profiled in the Places

Rated Almanac do not match the QOL issues that most concern local residents. They argue that

the data are of little use to planners because it does not monitor the smaller geographies within

metropolitan areas.

8. For particular life-course groups, the positive correlation may not hold. For example, when cor-

relating the in- and out-migration rates for highly educated young singles in metro areas, the

correlation coefficient is even slightly negative (r = -0.15, based on data taken from special

tabulations of the U.S. Census Bureau: Census 2000 PHC-T-34: Migration for the Young, Single,

and College-Educated for the United States, Regions, States, and Metropolitan Areas: 2000).

REFERENCESAdamson, D.W., D.E. Clark, and M.D. Partridge. 2004. Do urban agglomeration effects and household

amenities have a skill bias? Journal or Regional Science 44(2): 201–223.

Basker, E. 2002. Education, job search and migration. University of Missouri-Columbia Working

Paper 02-16. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=371120#PaperDownload

(accessed June 9, 2006).

Beale, C. 1969. The relationship of gross outmigration rates to net migration. Paper presented at the

annual meeting of the Population Association of America, Atlantic City, NJ, April 10–12.

Becker, G.S. 1993. Human capital, 3rd ed.. Chicago, IL: University of Chicago Press.

Bell, D. 1973. The coming of post-industrial society. New York: Basic Books.

Berry, B.J.L., and F.E. Horton. 1970. Geographic perspectives on urban systems. Englewood Cliffs,

NJ: Prentice-Hall.

Bloomquist, G.C., M.C. Berger, and J.P. Hoehn. 1988. New estimates of quality of life in urban areas.

American Economic Review 78(1): 89–107.

Borts, G., and J. Stein. 1964. Economic growth in a free market. NewYork: Columbia University Press.

Brown, L.A., and E.G. Moore. 1970. The intra-urban migration process: A perspective. Geografiska

Annaler 52B 1–13.

Carruthers, J.I., and B. Mundy, eds. 2006. Environmental valuation. Aldershot: Ashgate.

Clark, D., and W. Hunter. 1992. The impact of economic opportunity, amenities and fiscal factors on

age-specific migration rates. Journal of Regional Science 32: 349–365.


http://papers.ssrn.com/sol3/papers.cfm?abstract_id=371120#PaperDownload

Clark, D., T. Knapp, and N. White. 1996. Personal and location-specific characteristics and elderly

interstate migration. Growth and Change 27: 327–351.

Clark, T., R. Lloyd, K. Wong, and P Jain. 2002. Amenities drive urban growth. Journal of Urban

Affairs 24(5): 493–515.

Costa, D., and M. Kahn. 2000. Power couples: Changes in the locational choice of the college educated,

1940–1990. The Quarterly Journal of Economics 115(4): 1287–1315.

Cullen, J., and S. Levitt. 1999. Crime, urban flight, and the consequences for cities. The Review of

Economics and Statistics 81: 159–169.

Cushing, B. 1993. The effect of the social welfare system on metropolitan migration in the US, by

income group, gender and family structure. Urban Studies 30: 325–338.

Diamond, D.B., and G.S. Tolley, eds. 1982. The economics of urban amenities. New York: Academic

Press.

Dissart, J.-C., and S.C. Deller. 2000. Quality of life in the planning literature. Journal of Planning

Literature 1: 36–61.

Dugan, L. 1999. The Effect of victimization on a household’s moving decision. Criminology 37:

903–931.

Florida, R. 2002a. Bohemia and economic geography. Journal of Economic Geography 2: 55–

71.

———. 2002b. The rise of the creative class. New York: Basic Books.

Franklin, R. 2003. Migration of the young, single, and college educated: 1995 to 2000. Census 2000

Special Reports 1–11

Glaeser, E., J. Kolko, and A. Saiz. 2000. Consumer city. NBER Working Paper Series 7790.

Glaeser, E.L. 1999. Are cities dying? Journal of Economic Perspectives 12: 139–160.

Gleave, D., and M. Cordey-Hayes. 1977. Migration dynamics and labor market turnover. In Progress

in planning, Vol. 8, part 1, ed.. D. Diamond, and J. McLoughlin, 1–95. Oxford: Pergamon.

Golledge, R.G., and R.J. Stimson. 1997. Spatial behavior: A geographic perspective. New York: The

Guilford Press.

Gottlieb, P.D. 2001. The problem of Brain Drain in Ohio and Northeastern Ohio. What is it? How sever

is it? What should we do about it? Center for Regional Economic Issues. Case Western Reserve

University.

Gottlieb, P.D., and G. Joseph. 2006. College-to-work migration of technology graduates and holders of

doctorates within the United States. Journal of Regional Science 46(4): 627–649.

Graves, P. 1979. A life-cycle empirical analysis of migration and climate, by race. Journal of Urban

Economics 6: 135–147.

———. 1980. Migration and climate. Journal of Regional Science 20: 227–237.

———. 1983. Migration with a composite amenity: The role of rents. Journal of Regional Science 23:

541–546.

Graves, P., and P. Linneman. 1979. Household migration: Theoretical and empirical results. Journal of

Urban Economics 6: 383–404.

Greenwood, M.J. 1973. Urban economic growth and migration: Their interaction. Environment and

Behavior A 5: 91–112.

———. 1975. Research on migration in the United States: A survey. Journal of Economic Literature

13: 397–433.


Greenwood, M.J., G.L. Hunt, D.S. Rickman, and G.I. Treyz. 1991. Migration, regional equilibrium,

and the estimation of compensating differentials. American Economic Review 81: 1382–1390.

Greenwood, M.J., P.R. Mueser, D.A. Plane, and A.M. Schlottmann. 1991. New directions in migration

research: Perspectives from some North American regional science disciplines. Annals of Regional

Science 25: 237–270.

Gyourko, J., and J. Tracy. 1991. The structure of local public finance and the quality of life. Journal

of Political Economy 99: 774–806.

Hansen, S., C. Ban, and L. Huggins. 2003. Explaining the “brain Ddain” from older industrial cities:

The Pittsburgh Region. Economic Development Quarterly 17: 132–147.

Herzog, H., and A. Schlottmann. 1986. The metro rating game: What can be learned from recent

migrants? Growth and Change 17: 37–50.

Huffman, W.E., and T. Feridhanusetyawan. 2007. Migration, fixed costs, and location-specific ameni-

ties: A hazard analysis for a panel of males. American Journal of Agricultural Economics 89(2):

368–382.

Isserman, A.M. 1986. Population change and the economy: Social science theories and models.

Boston, MA: Kluwer Academic Publishers.

Jürges, H. 2006. Gender ideology, division of housework, and the geographic mobility of families.

Review of Economics of the Household 4(4): 299–323.

Kodrzycki, Y. 2001. Migration of recent college graduates: Evidence from the National Longitudinal

Survey of Youth. New England Economic Review 2001: 13–34.

Landis, J., and D. Sawicki. 1988. A planner’s guide to the Places Rated Almanac. Journal of the

American Planning Association 54: 336–346.

Levinson, D. 1978. The seasons of a man’s life. New York: Knopf.

Lowry, I. 1966. Migration and metropolitan growth: Two analytical models. San Francisco, CA:

Chandler.

Mathur, V. 1999. Human capital-based strategy for regional economic development. Economic Devel-

opment Quarterly 13(4): 203–221.

Mincer, J. 1978. Family migration decisions. Journal of Political Economy 86: 749–773.

Morenoff, J., and R. Sampson. 1997. Violent crime and the spatial dynamics of neighborhood

transition: Chicago, 1970–1990. Social Forces 76: 31–64.

Mueser, P., and P. Graves. 1995. Examining the role of economic opportunity and amenities in

explaining population redistribution. Journal of Urban Economics 37: 176–200.

Mulligan, G., J. Carruthers, and M. Cahill. 2004. Urban quality of life and public policy: A survey. In

Urban dynamics and growth 266, ed.. R. Capello and P. Nijkamp, 729–802. North Holland, The

Netherlands: Elsevier.

Nivalainen, S. 2004. Determinants of family migration: Short moves vs. long moves. Journal of

Population Economics 17(1): 157–175.

Pandit, K. 1997. Cohort and period effects in U.S. migration: How demographic and economic cycles

influence the migration schedule. Annals of the Association of American Geographers 87(3):

439–450.

Plane, D., and F. Heins. 2003. Age articulation of U.S. inter-metropolitan migration flows. Annals of

Regional Science 37: 107–130.


Plane, D., C. Henrie, and M. Perry. 2005. Migration up and down the urban hierarchy and across the

life course. Proceedings of the National Academy of Sciences 102: 15313–15318.

Plane, D., P. Rogerson, and A. Rosen. 1984. The cross-regional variation of in-migration and out-

migration. Geographical Analysis 16: 162–175.

Porell, F. 1982. Intermetropolitan migration and quality of life. Journal of Regional Science 22:

137–158.

Rappaport, J. 2003. Moving to nice weather. RWP 03-07. Federal Reserve Bank of Kansas City.

———. 2007. Moving to high quality of life. RWP 07-02. Federal Reserve Bank of Kansas City.

Roback, J. 1982. Wages, rents, and the quality of life. The Journal of Political Economy 90(6):

1257–1278.

Rogers, A. 1988. Age patterns of elderly migration: An international comparison. Demography 25(3):

355–370.

Rosen, S. 1974. Hedonic prices and implicit markets: Product differentiation in pure competition.

Journal of Political Economy 98: 1655–1666.

Rossi, P.H. 1955. Why families move: A study in the social psychology of urban residential mobility.

Glencoe, IL: Free Press.

Rowland, D.T. 1982. Living arrangements and the later family cycle in Australia. Australian Journal

of Ageing 1: 3–6.

Savageau, D., and G. Loftus. 1997. Places Rated Almanac, 5th ed. New York: Simon Schuster, Inc.

Schaeffer, P. 1987. A family model of migration. Socio-Economic Planning Sciences 21: 263–226.

Scott, A.J., ed.. 2001. Global city-regions: Trends, theory, policy. Oxford: Oxford University Press.

Sjaastad, L. 1962. The costs and returns of human migration. Journal of Political Economy 70(Suppl.):

80–93.

Smith, D.M. 1977. Human geography: A welfare approach. London: Edward Arnold.

Swain, L., and S. Garasky. 2007. Migration decisions of dual-earner families: An application of

multilevel modeling. Journal of Family and Economic Issues 28(1): 171–182.

Thirlwall, A.P. 1966. Migration and regional unemployment. Westminster Bank Review (November):

31–44.

Wakabayashi, M., and G.J.D. Hewings. 2007. Life-cycle changes in consumption behavior: Age-

specific and regional variations. Journal of Regional Science 47: 315–337.

Waldorf, B. 2006. Indiana’s knowledge corridor. Research Report, Purdue Center for Regional Devel-

opment, PCRD-R-1. http://www.ces.purdue.edu/extbusiness/pcrd/PCRD-R-1LR.pdf (accessed

August 3, 2006).

———. 2007. Brain drain in rural America. Selected Paper, Annual Meetings of the Association of

Agricultural Economics, Portland, OR, July 2007. http://agecon.lib.umn.edu/cgi-bin/pdf_view.pl?

paperid=26383&ftype=.pdf (accessed July 7, 2007).

Welch, R., J. Carruthers, and B. Waldorf. 2007. Public service expenditures as compensating differ-

entials in United States Metropolitan Areas: Housing values and rents. Cityscape: A Journal of

Policy Development and Research 9(1): 131–156.

Wilkinson, B., 2007. Leaving Los Angeles: Education and migration in the late-capitalist context. MA

Thesis, University of Arizona.

Wolpert, J. 1965. Behavioral aspects of the decision to migrate. Papers of the Regional Science

Association 15: 159–169.


http://www.ces.purdue.edu/extbusiness/pcrd/PCRD-R-1LR.pdf

http://agecon.lib.umn.edu/cgi-bin/pdf_view.pl?

quality of life and the migration of the college-educated ...naumannj/professional geography...

Documents